Modelpublic

jindaznb/torgo_tiny_finetune_M05_frozen_encoder

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes2downloads

Model Card

torgotinyfinetuneM05frozen_encoder

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.2755
—Wer: 40.5772

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

—learning_rate: 0.0001
—trainbatchsize: 16
—evalbatchsize: 1
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7762	0.84	500	0.2681	42.3599
0.0927	1.68	1000	0.2688	26.0611
0.0703	2.53	1500	0.2827	27.6740
0.0457	3.37	2000	0.2467	22.4109
0.0318	4.21	2500	0.2900	21.8166
0.0225	5.05	3000	0.2947	23.9389
0.0173	5.89	3500	0.2752	22.3260
0.0127	6.73	4000	0.2749	22.7504
0.0112	7.58	4500	0.2957	22.4109
0.008	8.42	5000	0.2765	23.3447
0.0071	9.26	5500	0.2780	30.3056
0.0049	10.1	6000	0.2827	23.5144
0.0045	10.94	6500	0.2884	34.5501
0.0036	11.78	7000	0.2605	36.1630
0.0028	12.63	7500	0.2787	30.5603
0.0024	13.47	8000	0.2758	31.5789
0.0016	14.31	8500	0.2801	33.1919
0.0018	15.15	9000	0.2779	33.9559
0.0011	15.99	9500	0.2737	37.2666
0.0008	16.84	10000	0.2757	31.5789
0.0005	17.68	10500	0.2787	35.6537
0.0004	18.52	11000	0.2747	35.9083
0.0003	19.36	11500	0.2755	40.5772

Framework versions

—Transformers 4.32.0
—Pytorch 2.1.0+cu121
—Datasets 2.14.7
—Tokenizers 0.13.3